atmospheric correction of high spatial resolution commercial satellite imagery products using modis...

Atmospheric Correction of High Spatial Resolution Commercial Satellite Imagery

Products Using MODIS Atmospheric Products

Mary Pagnutti

Science Systems and Applications, Inc.John C. Stennis Space Center, MS 39529

phone: 228-688-2135e-mail: mary.pagnutti@ssc.nasa.gov

3rd International Workshop on the Analysis of Multi-temporal Remote Sensing Images

May 17, 2005

Co-Authors / Contributors

NASA Applied Sciences Directorate, SSCTom Stanley* Vicki Zanoni

Science Systems and Applications, Inc.Slawomir Blonski Kara Holekamp*Kelly Knowlton Robert Ryan*

Computer Sciences CorporationJeffery A. Russell * Ronald D. Vaughan*Don Prados*

Lockheed Martin Space OperationsDavid Carver Jerry GasserRandy Greer Wes Tabor

* co-author

This work was directed by the NASA Applied Sciences Directorate (formerly the Earth Science Applications Directorate) at the John C. Stennis Space Center, Mississippi.

Participation in this work by Lockheed Martin Space Operations – Stennis Programs was supported under contract number NAS 13‑650. Participation in this work by Computer

Sciences Corporation and by Science Systems and Applications, Inc., was supported under NASA Task Order NNS04AB54T.

Overview

• Objective– Evaluate accuracy of a prototype algorithm that uses satellite-

derived atmospheric products to generate scene reflectance maps for high spatial resolution (HSR) systems

• Approach– Implement algorithm in an end-to-end process– Compare algorithm generated scene reflectance maps with

ground-truth data– Identify algorithm sensitivities– Provide recommendations

• Constraints– Ground truth available only in VNIR spectral range

Atmospheric Correction

• Atmospheric correction is the process of converting satellite signals (at-sensor radiance) to ground reflectance – Removes atmospheric and solar illumination effects

• Benefits– Improves change detection

– Used with spectral library based classifiers

– Simplifies satellite data intercomparisons

• Different levels of atmospheric correction yield different approximations of scene reflectance– Planetary reflectance – no knowledge of atmosphere

– Ground reflectance using knowledge of atmosphere

– Ground reflectance using knowledge of atmosphere and adjacency effects

Planetary Reflectance

First-order approximation – no knowledge of atmosphere

2

cos

d

EL sunp

TOA

distance Earth-Sunirradianceeric exoatmosph Solar

angle zenith Solarradiance sensor)-(at atmosphere of Top

ereflectancPlanetary :Where

dE

L

sun

TOA

p

Atmospheric Correction Algorithm Implementations

• Use knowledge of atmosphere to determine the constants necessary to convert satellite signals to scene reflectances– Ground-based reflectance measurements (direct method)– Pseudo-invariant targets – Ground-based atmosphere (aerosol) measurements– Scene-based aerosol estimates (based on dark pixels)– Climatological atmosphere– Satellite-based atmospheric measurements

Atmospheric Correction Prototype Algorithm

• Leverage JACIE commercial imagery radiometric characterizations– IKONOS, QuickBird, OrbView-3 (future)

• Use daily coverage from MODIS to provide input data for atmospheric correction– MOD04 Aerosol Optical Thickness– MOD05 Total Precipitable Water (Water Vapor)

• Generate MODIS-like products– Surface Reflectance (MOD09) – Gridded Vegetation Indices – NDVI (MOD13)

Water Vapor,Near Infrared

OpticalDepth

Atmospheric Correction Approach

MODIS data productsMOD04, MOD05

Spherical Albedo Model

Reflectance MapNDVI Map

Radiometrically Corrected Imagery

MODISModerate Resolution Imaging Spectroradiometer

VITAL FACTS:• Instrument: Whiskbroom imaging radiometer

• Bands: 36 from 0.4 and 14.5 µm

• Spatial Resolution: 250 m (2), 500 m (5), 1000 m (29)

• Swath: 2,300 km (55) from 705 km

• Repeat Time: Global coverage in 1 to 2 days

• Design Life: 6 years

MISSIONS:• Terra – Dec 1999

• Aqua – May 2002

MODIS provides long-term observations from which an enhanced knowledge of global dynamics and processes occurring on the surface of the Earth and in the lower atmosphere can be derived.

LINKS:• Sensor Site:

http://modis.gsfc.nasa.gov/

• Data Sites:http://daac.gsfc.nasa.gov/ (ocean and atmospheric)http://edcdaac.usgs.gov/main.html (land)

HERITAGE:• AVHRR

• High Resolution Infrared Radiation Sounder (HIRS)

• Landsat TM

• Coastal Zone Color Scanner

OWNER:• U.S., NASA

PRODUCT SUMMARY:• Congruent observations of high-priority

atmospheric, oceanic, and land-surface features

Spherical Albedo Formulation

The spherical albedo approach approximates the signal observed by the satellite as the summation of the components illustrated below

Target

SatelliteSolar Irradiance

Background

L0

Bbg

Atgt

s Atgtsbg

Bbgsbg

Bbgs2bg2

Atgts2bg2

Atmosphere with spherical albedo, s (aerosols, molecules,pressure, temperature, humidity)

...ρsBρsρBρBρ...ρsAρsρAρAρLL bgbgbgbgbgbgtgtbgtgttgtoTOA 2222

Atmospheric Correction Approximations

geometryandpropertiescatmospheri

ondependthatconstantsareand

radiance sensor)-(at atmosphere of Top

ereflectancBackground

ereflectanc Target

:Where

0

0

0

1

)(

11

LA, B, s,

L

s

BALL

s

B

s

ALL

TOA

bg

tgt

tgt

tgtTOA

bg

bg

bg

tgtTOA

Spherical Albedo formulation simplifies to:

Knowledge of atmosphere and adjacency

Knowledge of atmosphere

Adjacency Effects

• Adjacency effects are caused by complicated multiple scattering in the atmosphere-land surface interactions– Dark pixels appear brighter and bright pixels appear darker– Significant in turbid atmospheres over highly heterogeneous

landscapes

• Different methods have been employed for removing this effect– Atmospheric point spread function-PSF (Environmental

Function)– Empirical formula

Spherical Albedo Benefits

• Commonly used and found throughout the literature

• Allows for analytical determination of target albedo/reflectance values

• Computationally efficient

Atmospherically Corrected Imagery

IKONOS CIR image(rgb=431)

Stennis Space CenterJanuary 15, 2002

Includes material © Space Imaging, LLC

Atmospherically Corrected NDVI from IKONOS Imagery

Stennis Space CenterJanuary 15, 2002

Includes material © Space Imaging, LLC

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

NDVI from IKONOS Imagery

Stennis Space CenterJanuary 15, 2002

Includes material © Space Imaging, LLC

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 10

100

200

300

400

500

600

700

800

NDVI

Co

un

tPlanetary Reflectance

Radiance

Spher. Albedo

Spher. Albedo (Adj.)

Tarp Site: Stennis Space CenterJanuary 15, 2002

NDVI Histogram Comparison

Atmospheric Correction Prototype Algorithm Verification

Scene Selection for Atmospheric Correction Algorithm Verification

• Criteria– Available ground-truth reflectance and atmospheric

measurements– Available radiometric calibration

coefficients– MODIS overpass close in time to

IKONOS/QuickBird overpass

Selected Scene Matrix

Location Date HSR Sensor

SensorAz/El Ground Truth

SSC, MS Jan. 15, 2002 IKONOS 113.0 / 77.2

deg

Targets = 3 tarps (3.5, 22, 52), grass, concreteASD FieldSpec FR Spectroradiometer, ASR/MFRSR,

pressure, radiosonde, BRDF

SSC, MS Feb. 17, 2002 IKONOS 100.7 / 81.9

deg

Targets = 3 tarps (3.5, 22, 52), grass, concreteASD FieldSpec FR Spectroradiometer, ASR/MFRSR,

radiosonde, BRDF

SSC, MS Feb. 17, 2002 QuickBird 10.5 / 67.3

deg

Targets = 3 tarps (3.5, 22, 52), grass, concreteASD FieldSpec FR Spectroradiometer, ASR/MFRSR,

radiosonde, BRDF

Brookings, SD July 20, 2002 QuickBird 349.8 / 64.1

deg

Targets = 2 tarps (3.5, 52), grassASD FieldSpec FR Spectroradiometer, ASR/MFRSR,

radiosonde

Brookings, SD Sept. 7, 2002 QuickBird 191.0 / 74.9

degTargets = 2 tarps (3.5, 52), grass

ASD FieldSpec FR Spectroradiometer, ASR/MFRSR

SSC, MS Nov. 14, 2002 QuickBird 274.8 / 79.4

deg

Targets = 3 tarps (3.5, 22, 52), grass, concreteASD FieldSpec FR Spectroradiometer, ASR/MFRSR,

pressure, radiosonde, BRDF

Brookings, SD Aug. 23, 2003 QuickBird 148.3 / 76.8

degTargets = grass

ASD FieldSpec FR Spectroradiometer, ASR/MFRSR

Brookings, SD Oct. 21, 2003 QuickBird 279.5 / 81.3

degTargets = grass

ASD FieldSpec FR Spectroradiometer, ASR/MFRSR

SSC, MS Jan. 10, 2004 QuickBird 230.7 / 89.2

deg

Targets = 4 tarps (3.5, 22, 34, 52), grass, concreteASD FieldSpec FR Spectroradiometer, ASR/MFRSR,

pressure, radiosonde

Satellite Overpass Times

Location Date HSR SensorHSR Satellite

Overpass TimeMODIS Overpass

Time

SSC, MS Jan. 15, 2002 IKONOS 16:44 17:10

SSC, MS Feb. 17, 2002 IKONOS 16:47 16:14

SSC, MS Feb. 17, 2002 QuickBird 16:45 16:14

Brookings, SD July 20, 2002 QuickBird 17:26 17:42

Brookings, SD Sept. 7, 2002 QuickBird 17:22 16:47

SSC, MS Nov. 14, 2002 QuickBird 16:44 16:25

Brookings, SD Aug. 23, 2003 QuickBird 17:07 16:57

Brookings, SD Oct. 21, 2003 QuickBird 17:11 18:17

SSC, MS Jan. 10, 2004 QuickBird 16:30 17:27

Algorithm Validation using NASA/JACIE Generated Ground Truthing Datasets

• Laboratory Measurements– ASD FieldSpec FR Spectroradiometer calibrations – BRDF laboratory measurements

• Field Measurements– Radiometric calibration tarps, grass, and concrete targets– In-field calibrated sun photometers– In-field setup to check atmospheric model parameters

Calibration and Characterization of ASD FieldSpec Spectroradiometers

• NASA SSC maintains four ASD FieldSpec FR spectroradiometers– Laboratory transfer radiometers– Ground surface reflectance for

V&V field collection activities• Radiometric Calibration

– NIST-calibrated integrating sphere serves as source with known spectral radiance

• Spectral Calibration– Laser and pen lamp illumination of

integrating sphere• Environmental Testing

– Temperature stability tests performed in environmental chamber

Laboratory BRDF Measurements

• Purpose– Laboratory BRDF measurements

are used to correct ground-based reflectance measurements for satellite viewing and for solar illumination geometry

• Method– Collimated FEL lamp source– NIST-calibrated Spectralon® panel

serves as reference– Goniometer-mounted sample

controls illumination geometry– Optronics OL750 hyperspectral

instrument measures spectra

Wavelength (nm)

Ref

lect

ance

Algorithm Verification Case Matrix

• Three different atmospheric correction approximations– Case (1) Planetary reflectance– Case (2) Spherical albedo w/knowledge of atmosphere– Case (3) Spherical albedo w/knowledge of atmosphere &

adjacency• Three different sets of data used as input into approximation

– Case (a) ground based-sun photometer (aerosol), TOMS (ozone), Radiosonde (water vapor)

– Case (b) MOD04 (aerosol), TOMS (ozone), Radiosonde (water vapor)

– Case (c) MOD04 (aerosol), MOD05 (water vapor), TOMS (ozone) Operational Case

• Nine different scenes

Planetary Spherical Albedo Spherical Albedo w/adjacency

9 cases (1) + 17 cases (2b/2c) + 20 cases (3a/3b/3c) = 46 cases

MODIS/Sun Photometer Comparison

Total Optical DepthStennis Space Center

January 10, 2004

Reflectance Map Over SSC Tarps

January 10, 2004 – Case 3c ( Operational input of best approximation)

km

km

Blue Band Reflectance

0 0.2 0.4 0.6 0.8

0

0.2

0.4

0.6

0.8 0.1

0.2

0.3

0.4

0.5

km

km

Green Band Reflectance

0 0.2 0.4 0.6 0.8

0

0.2

0.4

0.6

0.8 0.1

0.2

0.3

0.4

0.5

km

km

Red Band Reflectance

0 0.2 0.4 0.6 0.8

0

0.2

0.4

0.6

0.8

0

0.1

0.2

0.3

0.4

0.5

km

km

NIR Band Reflectance

0 0.2 0.4 0.6 0.8

0

0.2

0.4

0.6

0.8

0

0.1

0.2

0.3

0.4

0.5

Measured and Calculated Reflectance Values of 52% Tarp

CASEreflectance

(Case reflectance – meas reflectance)BLUE GREEN RED NIR RMS

Measured Reflectance(Truth)

0.52 0.52 0.53 0.53 --

Case 1(Planetary reflectance)

0.46(-0.06)

0.46(-0.06)

0.48(-0.05)

0.50(-0.03)

0.05

Case 2c(Operational – no adjacency)

0.41(-0.11)

0.44(-0.08)

0.47(-0.06)

0.50(-0.03)

0.07

Case 3c(Operational – w/adjacency)

0.50(-0.02)

0.50(-0.02)

0.51(-0.02)

0.52(-0.01)

0.02

Important to take into account the adjacency effect

Algorithm Approximation Effects

Stennis Space CenterJanuary 10, 2004

Measured and Calculated Reflectance Values of Grass Target

CASEreflectance

(Case reflectance – meas reflectance)BLUE GREEN RED NIR RMS

Measured Reflectance(Truth)

0.05 0.07 0.10 0.18 --

Case 1(Planetary reflectance)

0.15(0.10)

0.13(0.06)

0.13(0.03)

0.20(0.02)

0.06

Case 2c(Operational – no adjacency)

0.06(0.01)

0.08(0.01)

0.11(0.01)

0.20(0.02)

0.01

Case 3c(Operational – w/adjacency)

0.06(0.01)

0.08(0.01)

0.11(0.01)

0.20(0.02)

0.01

No adjacency effect

Algorithm Approximation Effects

Stennis Space CenterJanuary 10, 2004

Measured and Calculated Reflectance Values of 52% Tarp

MODIS Input Parameter Effects

CASEreflectance

(Case reflectance – meas reflectance)BLUE GREEN RED NIR RMS

Measured Reflectance

(Truth)

0.52 0.52 0.53 0.53 --

Case 3a

(Meas. aerosol and water)

0.52(0.00)

0.52(0.00)

0.53(0.00)

0.53(0.00)

0.00

Case 3b

(MOD04 & meas. water)

0.50

(-0.02)

0.50

(-0.02)

0.51

(-0.02)

0.52

(-0.01)

0.02

Case 3c

(MOD04 & MOD05)

0.50

(-0.02)

0.50

(-0.02)

0.51

(-0.02)

0.52

(-0.01)

0.02

Stennis Space CenterJanuary 10, 2004

Measured and Calculated Reflectance Values using Operational Inputs

Stennis Space CenterJanuary 10, 2004

MOD04 – aerosolMOD05 – water vapor

CASE 3creflectance

(Case reflectance – meas reflectance)BLUE GREEN RED NIR RMS

3.5% Tarp(0.04, 0.04, 0.03, 0.03)

0.05(0.01)

0.06(0.02)

0.05(0.02)

0.06(0.03)

0.02

22% Tarp(0.22, 0.22, 0.21, 0.20)

0.24(0.02)

0.24(0.02)

0.23(0.02)

0.23(0.03)

0.02

34% Tarp(0.31, 0.31, 0.31, 0.30)

0.34(0.03)

0.34(0.03)

0.34(0.03)

0.34(0.04)

0.03

52% Tarp(0.52, 0.52, 0.53, 0.53)

0.50(-0.02)

0.50(-0.02)

0.51(-0.02)

0.52(-0.01)

0.02

Grass(0.05, 0.07, 0.10, 0.18)

0.06(0.01)

0.08(0.01)

0.11(0.01)

0.20(0.02)

0.01

Concrete(0.11, 0.13, 0.16, 0.19)

0.12(0.01)

0.15(0.02)

0.17(0.01)

0.21(0.02)

0.01

Operational Algorithm Verification Summary (Case 3c)

Location Date SensorAverage RMS for

all targets

SSC, MS Jan. 15, 2002 IKONOS 0.02

SSC, MS Feb. 17, 2002 IKONOS 0.04

SSC, MS Feb. 17, 2002 QuickBird 0.01

Brookings, SD July 20, 2002 QuickBird 0.01

Brookings, SD Sept. 7, 2002 QuickBird 0.02

Brookings, SD Aug. 23, 2003 QuickBird 0.00

Brookings, SD Oct. 21, 2003 QuickBird 0.02

SSC, MS Jan. 10, 2004 QuickBird 0.02

Summary/Future Direction

• MODIS products (MOD04, MOD05) provide the necessary inputs to generate high-spatial-resolution reflectance products under many conditions

– Average RMS differences range between 0.00–0.04 for the eight datasets evaluated (Case 3c-Operational input to best approximation)

• Adjacency can be an important component that needs to be accounted for to minimize errors

• Future Activities/Recommendations

– Evaluate alternate algorithms

– Compare algorithm results to MODIS products (MOD09, MOD13)

– Compare algorithm results to commercial atmospheric correction algorithm results (FLAASH, ACORN, ATCOR …)